Yeah but you always gotta think about the activities of those wanting to maximize and what they'd do. We tried at one point, can't remember what exactly the ruleset was, but if an account received the same vote too often it would grant diminishing returns (25% less value, followed by 50%, 75%, etc) and that mostly resulted in maximizers just create more accounts and voting on them there.
Times are different now of course so there may be some value behind this idea but it would require running some simulations of how it could potentially be abused and if it would cause any harm to authors.
I assume you mean the authors created more accounts for spam purposes (to hide their activities) and not the curators creating accounts. That has seemed to significantly go away and I would think the likelihood of that is the same if this change is made or not since self voting activities like that always have to operate in the dark.
But from just a straight “should I manual vote vs auto vote” perspective, I think it will overall come down to which gives a better return with minimal effort. Adding value to the manual vote with a layer two solution seems like a steep uphill battle. Reducing auto votes on the base layer seems like the only way and like you said, auto vote programs will just evolve, so detecting it has to be done another way. That is where my thinking was.
Current top authors will lose votes if people change their behavior, but if it is framed as properly rewarding under valued content and improving overall decentralization of the base token, I think the community can get behind that.